I've been lurking on this forum for half a decade now and figured it
was time I made a contribution ;-) I've found the various “search for
the holy upper ontology” discussions interesting – and John Sowa's
comment in this thread about what such an ontology might look like for
Watson caught my eye... (01)
> There is some discussion in the UMIA reports about a very simple
> ontology, which I suspect is not much more than a hierarchy of
> terms with very little axiomatization. (02)
This comment by Ed Barkmeyer on the earlier Watson thread also resonated: (03)
> In so many words, I think we can all agree that "the real Semantic Web"
> is going to be based on natural language processing -- a combination of
> smart algorithms and brute force -- and Watson is the demonstration that
> its time is now. (04)
Some of my company's recent experiences mirror both of these
“Watson”-related themes so I thought I'd share them. Please forgive
any formal inaccuracies in my comments here – I'm an applications and
database person (with some grounding in cognitive science), and not a
scholar in this field. (05)
I strongly believe that a universal “upper ontology” is possible if
its axioms are kept simple enough – a place where every “human”
concept exists or can exist without ambiguity. I'm happy to outline
my reasoning if anyone is interested – but really, the point is moot,
because we've gone ahead and built one anyway. (06)
The core tech for it is our own distributable graph database. Into
this we've plugged a (very) simple upper ontology based on WordNet.
It's so simple in fact that we're much more inclined to use the terms
“semantic map” or “universal schema”, depending on the audience. (07)
The question is: how useful is an ontology this simple? Well, not
very – if all you have is WordNet. So we threw in another 12 million
concepts extracted from dbPedia and located them within the map using
some basic NLP - to see what we could do with them. (08)
With little more than some optional directedness and quantification as
axioms it's difficult to do much sensible data extraction – let alone
reasoning. We've ended up with a sort of “semantic landscape” that is
really only useful if you're willing to process largish subgraphs.
We've had to rely heavily on the concepts surrounding those we're
interested in – their “context” if you will. This is where the “brute
force” comes in. A typical useful query requires hundreds of
thousands of traversals, most of which are four hops or more. It's not
the sort of approach you'd entertain without modern hardware. (09)
That said, we can do some pretty interesting things with it. So far,
this is what it looks like the map will be useful for... (010)
1) Connected, concept-based indexing and search. Even if that's all
it's good for, in many respects it's an improvement on the current
keyword paradigm.
2) A universal structure for shared data. If it can sensibly handle
“upper level” concepts, all other concepts and data are just a matter
of scale.
3) A framework for interoperability and upon which to hang a variety
of applications – including domain ontologies and other weak AI. i.e
a framework for doing a “Watson”. (011)
If it can be made truly universal in its usefulness, we believe the
map should be a community resource. I'm still not sure just what form
that “sharing” should take, but we've made a start by creating a new
take on the semantic wiki – you can see at on-line at http://wik.me.
It's an early beta, so please be gentle ;-) (012)
Some of the features of the wik.me site made possible by this semantic
map are... (013)
1) Self-organisation of facts in concept-focussed “pages”, based on
their position within the map.
2) Keyword search that finds concepts using their axioms.
3) The ability to modify the map in real-time using some basic NLP. (014)
What we're doing is not really "ontology" in the sense generally
discussed on this forum, but I'm happy to discuss it further if people
here are interested and think it's appropriate. Alternatively, there
are groups ( http://knowledgerights.org/group/wikmeTech and
http://knowledgerights.org/group/wikmeGovernance ) over at the
Knowledge Rights Forum that I've set up for the purpose of discussing
wik.me issues in detail. (015)
Steve (016)
On 22 February 2011 10:22, John F. Sowa <sowa@xxxxxxxxxxx> wrote:
> Ferenc,
>
> Since I did not work on Watson, all I know about it is what
> I have read about UMIA,... (017)
--
Stephen Young
CEO @ factnexus.com
Architect @ wik.me
Founding member @ knowledgerights.org (018)
_________________________________________________________________
Message Archives: http://ontolog.cim3.net/forum/ontolog-forum/
Config Subscr: http://ontolog.cim3.net/mailman/listinfo/ontolog-forum/
Unsubscribe: mailto:ontolog-forum-leave@xxxxxxxxxxxxxxxx
Shared Files: http://ontolog.cim3.net/file/
Community Wiki: http://ontolog.cim3.net/wiki/
To join: http://ontolog.cim3.net/cgi-bin/wiki.pl?WikiHomePage#nid1J
To Post: mailto:ontolog-forum@xxxxxxxxxxxxxxxx (019)
|